In general, when data analysis is executed by using a database appropriate for high-speed processing, and having a small capacity (for example, an in-memory database), analysis is executed by importing data necessary for data analysis from another database having a large capacity. In this case, there is the inconvenience of having to import data necessary for data analysis again when a data item different from the data item for which data analysis has already been executed is to be analyzed.
Further, when data (for example, meter data) to be periodically measured by a sensor is accumulated in a database having limited capacity, old data regarding which a fixed period of time has passed from acquisition of the data is moved to another database (for example, a tape device which is used for processing of a lower speed, and is low in price). In this case, there is the inconvenience of, when reference to the old data is desired, having to restore the old data from the tape device to the in-memory database.
In consideration of the inconveniences described above, there is a technique for accumulating necessary data in a database of fixed capacity by defining a significance threshold increasing with an elapse of time, and significance of data to be accumulated in a database by using a function of elapsed time, and storing only data items corresponding to a value greater than or equal to the significance threshold.
However, in the technique described above, there is the inconvenience of losing old data lower than the significance threshold by deletion with the passage of time, and thus there is the inconvenience of being unable to execute data analysis in which an influence of old data is taken into consideration.
A technical problem of the present invention is to provide a time-series data accumulation device capable of accumulating data in a form in which data analysis taking an influence of old data into consideration without deleting old data can be executed.